EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR CYBER-RESILIENT COMPUTING: AN ADAPTIVE FRAMEWORK FOR THREAT DETECTION AND RISK PREDICTION

The increasing complexity of cyber threats, coupled with the rapid expansion of cloud computing, Internet of Things (IoT) networks, and interconnected digital infrastructures, has created significant challenges for conventional cyber security mechanisms. Although artificial intelligence (AI) and machine learning (ML) have demonstrated considerable potential in identifying malicious activities, their limited interpretability, vulnerability to adversarial manipulation, and difficulties in adapting to evolving attack patterns restrict their effectiveness in security-critical environments. This study proposes an adaptive Explainable Artificial Intelligence (XAI) framework for cyber-resilient computing that integrates intelligent threat detection, interpretable decision-making, and dynamic risk prediction. The proposed architecture combines supervised and anomaly-based learning techniques with explainability mechanisms, including Shapley Additive explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME), to support transparent identification and interpretation of cyber threats. Furthermore, an adaptive risk assessment mechanism is introduced to prioritize detected threats according to predicted attack likelihood, potential impact, and system vulnerability. The framework is designed to accommodate changing network conditions through continuous monitoring, drift detection, and periodic model updates. A reproducible evaluation methodology is proposed using publicly available cyber security datasets, including CIC-IDS2017 and UNSW-NB15, with performance assessed through precision, recall, F1-score, false-positive rate, explanation stability, and risk prediction quality. Rather than claiming unverified experimental improvements, the study establishes a testable framework for investigating the relationship between detection accuracy, explainability, adaptability, and cyber resilience. The proposed approach aims to support accountable security decisions, proactive risk management, and more resilient computing infrastructures.